Region-adaptive concept aggregation for few-shot visual recognition
Few-shot learning (FSL) aims to learn novel concepts from very limited examples. However, most FSL methods suffer from the issue of lacking robustness in concept learning. Specifically, existing FSL methods usually ignore the diversity of region contents that may contain concept-irrelevant informati...
| Main Authors: | , , , , , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/169205 |
| _version_ | 1811680746002186240 |
|---|---|
| author | Han, Mengya Zhan, Yibing Yu, Baosheng Luo, Yong Hu, Han Du, Bo Wen, Yonggang Tao, Dacheng |
| author2 | School of Computer Science and Engineering |
| author_facet | School of Computer Science and Engineering Han, Mengya Zhan, Yibing Yu, Baosheng Luo, Yong Hu, Han Du, Bo Wen, Yonggang Tao, Dacheng |
| author_sort | Han, Mengya |
| collection | NTU |
| description | Few-shot learning (FSL) aims to learn novel concepts from very limited examples. However, most FSL methods suffer from the issue of lacking robustness in concept learning. Specifically, existing FSL methods usually ignore the diversity of region contents that may contain concept-irrelevant information such as the background, which would introduce bias/noise and degrade the performance of conceptual representation learning. To address the above-mentioned issue, we propose a novel metric-based FSL method termed region-adaptive concept aggregation network or RCA-Net. Specifically, we devise a region-adaptive concept aggregator (RCA) to model the relationships of different regions and capture the conceptual information in different regions, which are then integrated in a weighted average manner to obtain the conceptual representation. Consequently, robust concept learning can be achieved by focusing more on the concept-relevant information and less on the conceptual-irrelevant information. We perform extensive experiments on three popular visual recognition benchmarks to demonstrate the superiority of RCA-Net for robust few-shot learning. In particular, on the Caltech-UCSD Birds-200-2011 (CUB200) dataset, the proposed RCA-Net significantly improves 1-shot accuracy from 74.76% to 78.03% and 5-shot accuracy from 86.84% to 89.83% compared with the most competitive counterpart. |
| first_indexed | 2024-10-01T03:29:56Z |
| format | Journal Article |
| id | ntu-10356/169205 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:29:56Z |
| publishDate | 2023 |
| record_format | dspace |
| spelling | ntu-10356/1692052023-07-06T07:23:39Z Region-adaptive concept aggregation for few-shot visual recognition Han, Mengya Zhan, Yibing Yu, Baosheng Luo, Yong Hu, Han Du, Bo Wen, Yonggang Tao, Dacheng School of Computer Science and Engineering Engineering::Computer science and engineering Concept-Aggregation Concept Learning Few-shot learning (FSL) aims to learn novel concepts from very limited examples. However, most FSL methods suffer from the issue of lacking robustness in concept learning. Specifically, existing FSL methods usually ignore the diversity of region contents that may contain concept-irrelevant information such as the background, which would introduce bias/noise and degrade the performance of conceptual representation learning. To address the above-mentioned issue, we propose a novel metric-based FSL method termed region-adaptive concept aggregation network or RCA-Net. Specifically, we devise a region-adaptive concept aggregator (RCA) to model the relationships of different regions and capture the conceptual information in different regions, which are then integrated in a weighted average manner to obtain the conceptual representation. Consequently, robust concept learning can be achieved by focusing more on the concept-relevant information and less on the conceptual-irrelevant information. We perform extensive experiments on three popular visual recognition benchmarks to demonstrate the superiority of RCA-Net for robust few-shot learning. In particular, on the Caltech-UCSD Birds-200-2011 (CUB200) dataset, the proposed RCA-Net significantly improves 1-shot accuracy from 74.76% to 78.03% and 5-shot accuracy from 86.84% to 89.83% compared with the most competitive counterpart. This work was supported by National Natural Science Foundation of China (No. 62002090), Major Science and Technology Innovation 2030 “New Generation Artificial Intelligence” Key Project (No. 2021ZD0111700) and Special Fund of Hubei Luojia Laboratory, China (No. 220100014). 2023-07-06T07:23:38Z 2023-07-06T07:23:38Z 2023 Journal Article Han, M., Zhan, Y., Yu, B., Luo, Y., Hu, H., Du, B., Wen, Y. & Tao, D. (2023). Region-adaptive concept aggregation for few-shot visual recognition. Machine Intelligence Research. https://dx.doi.org/10.1007/s11633-022-1358-8 2731-538X https://hdl.handle.net/10356/169205 10.1007/s11633-022-1358-8 2-s2.0-85149140622 en Machine Intelligence Research © 2023 Institute of Automation, Chinese Academy of Sciences and Springer-Verlag GmbH Germany, part of Springer Nature. All rights reserved. |
| spellingShingle | Engineering::Computer science and engineering Concept-Aggregation Concept Learning Han, Mengya Zhan, Yibing Yu, Baosheng Luo, Yong Hu, Han Du, Bo Wen, Yonggang Tao, Dacheng Region-adaptive concept aggregation for few-shot visual recognition |
| title | Region-adaptive concept aggregation for few-shot visual recognition |
| title_full | Region-adaptive concept aggregation for few-shot visual recognition |
| title_fullStr | Region-adaptive concept aggregation for few-shot visual recognition |
| title_full_unstemmed | Region-adaptive concept aggregation for few-shot visual recognition |
| title_short | Region-adaptive concept aggregation for few-shot visual recognition |
| title_sort | region adaptive concept aggregation for few shot visual recognition |
| topic | Engineering::Computer science and engineering Concept-Aggregation Concept Learning |
| url | https://hdl.handle.net/10356/169205 |
| work_keys_str_mv | AT hanmengya regionadaptiveconceptaggregationforfewshotvisualrecognition AT zhanyibing regionadaptiveconceptaggregationforfewshotvisualrecognition AT yubaosheng regionadaptiveconceptaggregationforfewshotvisualrecognition AT luoyong regionadaptiveconceptaggregationforfewshotvisualrecognition AT huhan regionadaptiveconceptaggregationforfewshotvisualrecognition AT dubo regionadaptiveconceptaggregationforfewshotvisualrecognition AT wenyonggang regionadaptiveconceptaggregationforfewshotvisualrecognition AT taodacheng regionadaptiveconceptaggregationforfewshotvisualrecognition |